Help Needed for Oil Droplet Identification

[MembraneSci]

Hi folks,

I'm looking for help on identifying the droplets in three photos (below signature). I've tried many different combinations of commands (subtract background, threshold, gaussian blur, find edges, sharpen, enhance contrast, make binary, and watershed), but I just can't seem to get a good enough threshold image for accurate particle sizing.

I would be so appreciative to see a list of commands that can get me a good image!

Thanks for your help,
Leo

[Christine Labno-2]

Hi Leo,

Here's what I worked out on the middle file:

run("8-bit");
run("Subtract Background...", "rolling=2");
run("Gaussian Blur...", "sigma=1");
//run("Threshold...");
setAutoThreshold("Default");
setThreshold(0, 1);  
// do NOT check "dark background" here
// yes, you will be thresholding the "background" and not the outline.  
// The idea is to measure the negative space
// within the illuminated droplet outlines
// This way the particles are separated without watershed
run("Convert to Mask");
run("Dilate");
run("Analyze Particles...", "  circularity=0.60-1.00 show=[Bare Outlines] display exclude summarize");

It's far from perfect, so it will need modification, but I can pick up every particle that isn't touching the edge of the image and get close to the right size measurement.  

The main challenge that I see is your light source.  Since the droplets are illuminated from one side, there is a "thicker" side and a "thinner" side to the outline of the droplets.  Subtracting background with a very small rolling ball helps minimize this, but I can't get rid of it completely, so the resulting outlines are all a little bit short.  I tried color threshold and psedu-flatfield correction and neither of those things is able to eliminate the effect of the uneven lighting, but someone else may have an idea that will get you closer to perfect measurement.

Best,
Christine  

--------------------------------------------
Christine Labno, Ph.D.
Asst. Technical Director
Light Microscopy Core
University of Chicago
Office of Shared Research Facilities
KCBD 1250 900 E. 57th St.
[phone] (phone)

________________________________________
From: ImageJ Interest Group [[hidden email]] on behalf of MembraneSci [[hidden email]]
Sent: Thursday, April 21, 2016 4:03 PM
To: [hidden email]
Subject: Help Needed for Oil Droplet Identification

Hi folks,

I'm looking for help on identifying the droplets in three photos (below
signature). I've tried many different combinations of commands (subtract
background, threshold, gaussian blur, find edges, sharpen, enhance contrast,
make binary, and watershed), but I just can't seem to get a good enough
threshold image for accurate particle sizing.

I would be so appreciative to see a list of commands that can get me a good
image!

Thanks for your help,
Leo

<http://imagej.1557.x6.nabble.com/file/n5016212/20_80_Hex_in_Water_-_feed.bmp>
<http://imagej.1557.x6.nabble.com/file/n5016212/50_50_Hex_in_Water_-_feed.bmp>
<http://imagej.1557.x6.nabble.com/file/n5016212/80_20_H_W_tr_feed.bmp>



--
View this message in context: http://imagej.1557.x6.nabble.com/Help-Needed-for-Oil-Droplet-Identification-tp5016212.html
Sent from the ImageJ mailing list archive at Nabble.com.

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Philip Ershler-2]

Hello,
   Although I don't know much about it, I think what you are looking for is the Hough Transform. If you goggle for "ImageJ find circles", it may get you started.

Good Luck,
Phil

> On Apr 21, 2016, at 4:46 PM, Christine Labno <[hidden email]> wrote:
>
> Hi Leo,
>
> Here's what I worked out on the middle file:
>
> run("8-bit");
> run("Subtract Background...", "rolling=2");
> run("Gaussian Blur...", "sigma=1");
> //run("Threshold...");
> setAutoThreshold("Default");
> setThreshold(0, 1);  
> // do NOT check "dark background" here
> // yes, you will be thresholding the "background" and not the outline.  
> // The idea is to measure the negative space
> // within the illuminated droplet outlines
> // This way the particles are separated without watershed
> run("Convert to Mask");
> run("Dilate");
> run("Analyze Particles...", "  circularity=0.60-1.00 show=[Bare Outlines] display exclude summarize");
>
> It's far from perfect, so it will need modification, but I can pick up every particle that isn't touching the edge of the image and get close to the right size measurement.  
>
> The main challenge that I see is your light source.  Since the droplets are illuminated from one side, there is a "thicker" side and a "thinner" side to the outline of the droplets.  Subtracting background with a very small rolling ball helps minimize this, but I can't get rid of it completely, so the resulting outlines are all a little bit short.  I tried color threshold and psedu-flatfield correction and neither of those things is able to eliminate the effect of the uneven lighting, but someone else may have an idea that will get you closer to perfect measurement.
>
> Best,
> Christine  
>
> --------------------------------------------
> Christine Labno, Ph.D.
> Asst. Technical Director
> Light Microscopy Core
> University of Chicago
> Office of Shared Research Facilities
> KCBD 1250 900 E. 57th St.
> [phone] (phone)
>
> ________________________________________
> From: ImageJ Interest Group [[hidden email]] on behalf of MembraneSci [[hidden email]]
> Sent: Thursday, April 21, 2016 4:03 PM
> To: [hidden email]
> Subject: Help Needed for Oil Droplet Identification
>
> Hi folks,
>
> I'm looking for help on identifying the droplets in three photos (below
> signature). I've tried many different combinations of commands (subtract
> background, threshold, gaussian blur, find edges, sharpen, enhance contrast,
> make binary, and watershed), but I just can't seem to get a good enough
> threshold image for accurate particle sizing.
>
> I would be so appreciative to see a list of commands that can get me a good
> image!
>
> Thanks for your help,
> Leo
>
> <http://imagej.1557.x6.nabble.com/file/n5016212/20_80_Hex_in_Water_-_feed.bmp>
> <http://imagej.1557.x6.nabble.com/file/n5016212/50_50_Hex_in_Water_-_feed.bmp>
> <http://imagej.1557.x6.nabble.com/file/n5016212/80_20_H_W_tr_feed.bmp>
>
>
>
> --
> View this message in context: http://imagej.1557.x6.nabble.com/Help-Needed-for-Oil-Droplet-Identification-tp5016212.html
> Sent from the ImageJ mailing list archive at Nabble.com.
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[MembraneSci]

Hi Christine,

Thanks for your help on this. I'm still hoping to find an algorithm that will capture the whole circle instead of part, but this code is a good start for me.

Thanks again,
Leo